Remove redundant if clause and update documentation

src/adapters/methods/modeling.py

@@ -661,8 +661,8 @@ def _init_W(self, W_left=None, W_right=None, W=None):
 
     def reset_parameters(self):
 
-        if self.config.init_weights_seed:
-            fix_seed(self.config.init_weights_seed)
+        # Set seed for reproducibility if specified in config
+        fix_seed(self.config.init_weights_seed)
 
         if not self.shared_W_phm:
             self._init_W()

src/adapters/methods/prefix_tuning.py

@@ -31,8 +31,8 @@ def __init__(
         self.n_embd_per_head = n_embd_per_head or self.input_size // self.n_heads
         self.config = config
 
-        if self.config.init_weights_seed:
-            fix_seed(self.config.init_weights_seed)
+        # Set seed for reproducibility if specified in config
+        fix_seed(self.config.init_weights_seed)
         self.wte = nn.Embedding(self.config.prefix_length, self.input_size)
         self.control_trans = nn.Sequential(
             nn.Linear(self.input_size, self.config.bottleneck_size),
@@ -83,8 +83,8 @@ def __init__(
         self.n_embd_per_head = n_embd_per_head or self.input_size // self.n_heads
         self.config = config
 
-        if self.config.init_weights_seed:
-            fix_seed(self.config.init_weights_seed)
+        # Set seed for reproducibility if specified in config
+        fix_seed(self.config.init_weights_seed)
 
         self.control_trans = nn.Parameter(
             torch.randn(self.config.prefix_length * self.n_layers * 2 * self.n_heads * self.n_embd_per_head)

src/adapters/methods/prompt_tuning.py

@@ -67,8 +67,8 @@ def __init__(
 
     def _init_prompt_embedding(self, base_model_embeddings: nn.Module) -> None:
 
-        if self.prompt_tuning_config.init_weights_seed:
-            fix_seed(self.prompt_tuning_config.init_weights_seed)
+        # Set seed for reproducibility if specified in config
+        fix_seed(self.prompt_tuning_config.init_weights_seed)
 
         if self.prompt_tuning_config.prompt_init == "random_uniform":
             nn.init.uniform_(

src/adapters/methods/reft.py

@@ -29,8 +29,8 @@ def __init__(
         super().__init__()
         self.orthogonal = orthogonal
 
-        if init_weights_seed:
-            fix_seed(init_weights_seed)
+        # Set seed for reproducibility if specified in config
+        fix_seed(init_weights_seed)
         self.learned_source = nn.Linear(in_dim, r_dim, bias=True, dtype=dtype)
         projection = nn.Linear(in_dim, r_dim, bias=False, dtype=dtype)
 

src/adapters/methods/utils.py

@@ -20,6 +20,7 @@
 def fix_seed(seed: Optional[int] = None):
     """
     Helper function to fix the torch seed on cpu and gpu for initializing adapters with the same weights.
+    Is only executed if the config provides a respective seed.
     """
     if seed:
         torch.manual_seed(seed)